Accessibility

The Trouble with Accessibility Overlays

Overlays promise instant WCAG compliance from a single line of JavaScript. After auditing dozens of sites that bought that promise, the picture is far less reassuring — and the real fix is more ordinary than any widget will admit.

A circular diagram of accessibility icons including glasses, text-to-speech, magnifier, captions and touch
Real accessibility is built into content and code, not bolted on at the end.
Written by Sofia Reyes, Accessibility & Content Specialist Independently reviewed and fact-checked Last updated Jun 5, 2026 4 sources cited

Key takeaways

  • An accessibility overlay is a third-party JavaScript widget that patches a finished page at runtime — it cannot rewrite the source it sits on top of.
  • Overlays address only a small fraction of the WCAG success criteria, and several of their automated "fixes" introduce new barriers for screen reader and keyboard users.
  • Installing an overlay does not make a site WCAG compliant and has not reliably stopped lawsuits — many demand letters target overlay-equipped sites.
  • Real accessibility flows from the POUR principles: perceivable, operable, understandable and robust content built into the HTML, the design and the editorial process.
  • Durable remediation means semantic markup, real testing with assistive technology, and accessibility checks built into your everyday workflow.

Somewhere on the internet right now, a small organisation is reading a message that promises to make their website fully compliant with accessibility law in minutes. No developers, no audit, no rework — just paste one line of JavaScript and a tidy floating icon appears in the corner of every page. The pitch is seductive precisely because the alternative sounds expensive and slow. If a single script can guarantee compliance and ward off lawsuits, who would not take that deal?

The trouble is that the deal does not hold. Accessibility is not a coat of paint you can apply to a finished building; it is closer to the plumbing and the load-bearing walls. The most common accessibility overlays problems stem from a simple architectural mismatch: the barriers people with disabilities encounter live in the structure, content and behaviour of a page, and an overlay runs after all of that has already been decided. It can poke at the surface, but it cannot reach the foundations.

This article explains what overlays actually do, why they fall short of their headline promise, what the legal landscape really looks like, and — most importantly — what genuine web accessibility remediation involves. The reassuring conclusion at the end is that the real work is more ordinary, more durable and more within reach than the widgets would have you believe. The capability is almost certainly already inside your team.

What accessibility overlays promise

The marketing for overlay products tends to make three big claims. First, instant compliance: install the script and your site meets the Web Content Accessibility Guidelines (WCAG) without any other changes. Second, automation: artificial intelligence scans your pages and fixes whatever is broken, continuously, with no human involvement. Third, legal protection: with the widget in place, you are shielded from the wave of accessibility lawsuits and demand letters that have become common in recent years.

Each claim is appealing, and each is, at best, a serious oversimplification. The question of whether accessibility overlays are WCAG compliant is not a matter of opinion — it is testable, and the answer is consistently that an overlay alone does not produce a conformant page. The phrase "AI fixes everything" papers over the fact that many accessibility problems require human judgement about meaning and context that no scanner can supply. And the legal-shield claim, as we will see, has not survived contact with actual courtrooms.

It is worth being fair to the people who buy these tools. They are usually acting in good faith, often after a frightening demand letter, and they have been told by a vendor they trust that the problem is solved. The failure here is not the customer's; it is the gap between what is promised and what software running in the browser can physically do. That gap is wide, and the rest of this piece is about why.

How overlays actually work

To understand the limits, you have to understand the mechanism. An overlay is a piece of JavaScript hosted by a third party and loaded into your page. When a visitor's browser renders your site, the overlay script runs and does two broad things. It adds a toolbar — usually a floating button that opens a panel of toggles for larger text, higher contrast, a "dyslexia-friendly" font, a reading guide and so on. And it attempts automated remediation: scanning the rendered document and modifying it on the fly, injecting ARIA roles and patterns onto elements, guessing at labels for unlabelled buttons, or rewriting attributes it judges to be wrong.

Two features of that design matter enormously. The first is that everything happens at runtime, in the visitor's browser, after the page has loaded. The overlay never touches your source code; it manipulates a copy of the page in memory each time someone visits. The second is that the remediation is inferential. The script does not know what your image depicts or what your unlabelled icon does; it makes a statistical guess. When the guess is right, the result can be a small improvement. When it is wrong — and it is wrong often enough to matter — it confidently announces incorrect information to people who have no way to know it is incorrect.

From the field. When we tested a well-known overlay on a client's checkout flow using a screen reader, the widget had helpfully relabelled the "Remove item" button as simply "Button". A blind tester clicked it expecting to expand a section and instead emptied part of their cart. The page passed the overlay's own automated scan. A pattern we keep seeing is that overlays optimise for the metrics a robot can check, not for whether a human can actually complete the task.

Why one line of JavaScript cannot fix accessibility

Here is the heart of why accessibility overlays do not work as advertised. Accessibility, as codified in WCAG, is a set of dozens of testable success criteria. Independent analysis of automated tooling has long suggested that machines can reliably detect only a minority of those criteria, and they can confidently fix even fewer. The rest require understanding what content means — and that is exactly what an after-the-fact script cannot do.

Consider alternative text for images. The criterion is not "every image has an alt attribute"; it is "the alternative text conveys the same information or function as the image." An overlay can detect a missing alt and may slap on a machine-generated caption, but it cannot know that the photo on your landing page is decorative and should be silent, or that the chart is the entire point of the paragraph and needs a careful description. Meaning is human. The same is true of reading order, error messages that actually explain how to fix the problem, link text that makes sense out of context, and heading structures that mirror the logic of the content.

New barriers, not just missed ones

If overlays merely failed to fix things, they would be a waste of money but mostly harmless. The deeper concern is that they can actively interfere. Many assistive technologies are sophisticated, finely tuned tools, and their users have configured them carefully. When an overlay injects its own keyboard handlers, focus traps or ARIA, it can fight the screen reader the visitor already relies on — double-announcing content, stealing focus, breaking shortcuts, or overriding the operating system's own accessibility settings.

Watch for this. Some overlays intercept the keyboard and impose their own navigation model. For a sighted mouse user this is invisible, so it sails through internal testing. For a keyboard-only or screen reader user it can make a previously usable page worse than it was before the widget was installed — which is the exact opposite of remediation.

There is also a practical performance cost. An overlay is render-blocking third-party code that loads on every page, and it works against the kind of fast, stable experience captured by Core Web Vitals. Accessibility and performance are allies, not rivals; a page that is quick, predictable and built from clean semantic HTML tends to be more accessible by default. Bolting on a heavy script that rewrites the DOM on every visit pushes in the wrong direction on both fronts. If you want a grounding in how clean structure underpins everything else, our explainer on web development fundamentals, clearly explained is a useful companion to this piece.

The legal-protection claim deserves its own scrutiny, because it is the one that pushes the most people to buy. In the United States, the relevant law for most public-facing sites is the Americans with Disabilities Act. As the guidance on ADA.gov makes clear, the standard a court cares about is whether people with disabilities can actually access and use your services — not whether you have installed any particular tool.

That distinction is fatal to the shield argument. A widget that does not make the underlying experience usable does not satisfy the law simply by existing. In practice, the opposite has frequently happened: large numbers of demand letters and lawsuits have named sites that were already running overlays at the time of the complaint. In some cases plaintiffs have specifically argued that the overlay itself created or worsened barriers. A floating accessibility icon is, if anything, a flag that advertises your site as a target while doing little to remove the substance of the complaint.

What actually helps legally. Courts and regulators respond to evidence of real conformance: an audit against WCAG, a remediation plan with dates, testing that includes assistive technology, and an accessibility statement describing your standard and how users can report problems. Documentation of genuine effort and progress is worth far more than any badge a script can render.

None of this is legal advice, and the details vary by jurisdiction and by the kind of organisation you are. But the strategic point is durable across all of them: the safest position is a site that genuinely works for disabled users, evidenced by testing, not a site that merely looks compliant to a non-disabled buyer. The same caution applies to any "AI solves it" promise — a theme we return to in our look at AI in 2026: what every developer should actually know.

What real accessibility looks like (POUR and WCAG)

If overlays are the wrong answer, what is the right one? It starts with the framework that WCAG is built on, summarised by four principles known by the acronym POUR. The W3C Web Accessibility Initiative maintains these guidelines, and they are less intimidating than their reputation once you see the shape of them.

Perceivable means people can perceive the content through whatever sense they are using — text alternatives for images, captions for video, sufficient colour contrast, and content that does not rely on colour alone to convey meaning. Operable means everyone can drive the interface — full keyboard operability, no traps, a visible focus indicator, enough time to complete tasks, and no content that flashes in ways that can trigger seizures. Understandable means the content and behaviour are predictable — clear language, consistent navigation, labelled form fields, and error messages that say what went wrong and how to fix it. Robust means the markup is clean enough that current and future assistive technologies can interpret it reliably — which is mostly a matter of valid, semantic HTML and ARIA used correctly, or not at all.

Notice what is common to all four: they describe properties of the page's own structure and content, decided by the people who build and write it. This is why accessibility belongs in the same conversation as the rest of your craft — the kind of end-to-end thinking covered in our complete guide to modern web development and design. It also explains why a content system matters: the way authors enter headings, alt text and links every day determines accessibility at scale, a theme we explore in six real benefits of a modern CMS.

Overlay versus real remediation, side by side

The contrast is clearest in a table. The left column is the promise; the right column is what the work actually requires.

Dimension Accessibility overlay Real remediation
Where the change happens At runtime, in each visitor's browser In the source HTML, CSS, content and templates
Coverage of WCAG criteria A small, surface-level subset The full set, including human-judgement criteria
Permanence Temporary; gone if the script fails to load Permanent; baked into the codebase
Effect on assistive tech Can conflict with the user's own tools Works with every assistive technology by design
Performance impact Extra render-blocking third-party script Often improves speed and stability
Legal standing Frequently named in lawsuits anyway Evidence of genuine, testable conformance
Who benefits Mainly the buyer's sense of having "done something" The disabled users the law is actually about

Read down the right column and a reassuring truth emerges: none of it is exotic. It is disciplined, ordinary engineering and editorial care, applied consistently. That is harder to package as a one-line script, but it is the only thing that delivers.

A practical remediation roadmap

So how do you make a website accessible for real? Here is the roadmap I walk teams through, ordered roughly by sequence and by impact.

1. Audit honestly

Start by finding out where you actually stand. Run an automated checker to catch the easy, machine-detectable issues — missing labels, contrast failures, empty links — but treat that as the floor, not the ceiling. Then do manual testing: navigate the whole site with only the keyboard, and listen to your key pages with a screen reader. The gap between what the automated tool reports and what you experience by hand is, in effect, the gap an overlay can never close.

2. Fix the structure first

Most issues trace back to markup. Use real headings in logical order, not styled div elements. Use buttons for actions and links for navigation. Label every form field and associate every error with its input. Give every meaningful image accurate alternative text and let decorative images be silent. Ensure a visible focus indicator and a sensible focus order. These structural fixes resolve a large share of findings at once and make everything downstream easier.

3. Use ARIA sparingly and correctly

ARIA is powerful and dangerous in equal measure; the first rule is to prefer native HTML elements that already carry the right semantics. When you do need ARIA — for a custom widget, say — follow the established patterns in the ARIA Authoring Practices Guide rather than improvising. Incorrect ARIA is worse than none, because it actively lies to assistive technology, which is precisely the failure mode overlays fall into at scale.

4. Test with real people and real tools

Automated and manual testing by your own team is essential but incomplete. Nothing substitutes for watching people who actually use assistive technology attempt real tasks on your site. Their feedback surfaces issues no checklist anticipates and reorders your priorities around what matters. AI tooling can assist your checks here, but it does not replace lived experience — a boundary we examine in our practitioner take on putting AI at the core of your stack, carefully.

5. Make it continuous

Accessibility decays the moment you stop paying attention; the next deploy or the next article can reintroduce barriers. Bake checks into your build pipeline, add accessibility to code review and your definition of done, and train content authors on the few habits that matter most — headings, alt text, link text and tables. Choosing the right platform and tooling to support that habit is itself a decision worth making deliberately, as we discuss in how to choose the right software solution. Publish an accessibility statement so users know your standard and have a way to report problems.

When, if ever, overlays make sense

It would be dishonest to claim every feature in every overlay product is worthless. Some of the user-facing preference controls — a contrast toggle, a text-resize control, a reading guide — can be mildly useful conveniences for some visitors. The important caveat is that these are not fixes for an inaccessible site, and most of them duplicate capabilities the browser and operating system already provide more reliably. A genuinely accessible page lets a user bring their own tools and settings without needing a vendor's panel at all.

If you have inherited an overlay, the pragmatic path is not necessarily to rip it out in a panic, but to stop treating it as your accessibility strategy. Begin the real remediation work behind it, test whether the widget is helping or actively interfering with assistive technology, and plan to retire it once the underlying site can stand on its own. The destination is a page that needs no overlay because the structure underneath is sound.

The larger lesson is one that recurs across technology: there is rarely a single line of code that absolves you of doing the actual work. Accessibility is a quality of how you build and write, sustained over time, not a product you install. That is genuinely good news. It means the capability is already inside your team — in your developers' command of semantic markup and your authors' care with content — rather than locked behind a subscription to a script. If you want to keep going, browse the full library of articles or read more about who writes this journal on the about page; for a step back to first principles, our look at what exactly a software solution is frames why "buy a widget" so often misreads the problem.

Build it in. Test it with the people it is for. Keep it honest. That is the whole of the method, and no overlay can shortcut it.

Frequently asked questions

Do accessibility overlays make my site WCAG compliant?

No. An overlay cannot deliver full WCAG conformance because most success criteria depend on the underlying HTML, content and design — things a script layered on top cannot reliably rewrite. Overlays may patch a few surface issues, but genuine compliance requires changes in the source code, the content and the authoring process behind it.

Can an accessibility overlay protect me from a lawsuit?

Not dependably. Many lawsuits and demand letters have named sites that already ran overlays, and some plaintiffs argue the widget itself created barriers. The ADA judges whether disabled users can actually use your site, not whether you installed a tool. Real conformance, documented and tested, is the only durable defence.

What is the difference between an overlay and real remediation?

An overlay is a third-party script that tries to patch a finished page at runtime in the browser. Real remediation changes the source: the HTML structure, ARIA, contrast, labels, keyboard handling and content. Remediation is permanent, testable and benefits every assistive technology; an overlay is temporary and only as good as its guesswork.

How do I actually make my website accessible?

Start from WCAG and the POUR principles. Use semantic HTML, label every control, ensure keyboard operability and sufficient contrast, and add meaningful alternative text. Then test with real screen readers and keyboard-only navigation, involve people with disabilities, and bake accessibility checks into your build and authoring workflow so regressions are caught early.

Sources & further reading

  1. W3C — Web Content Accessibility Guidelines (WCAG) — the international standard defining the testable success criteria for accessible web content.
  2. W3C Web Accessibility Initiative (WAI) — the W3C group that develops accessibility guidelines, the POUR principles and supporting techniques.
  3. ARIA Authoring Practices Guide — patterns and worked examples for using ARIA roles, states and properties correctly.
  4. ADA.gov — official U.S. Department of Justice information on the Americans with Disabilities Act and web accessibility expectations.